fast-math

auto vectorization on double and ffast-math

Why is mandatory to use -ffast-math with g++ to achieve vectorization of loop using doubles? I don't like ffast-math because I don't want to loose precision. ...